SpaceX and xAI join Pentagon contest for voice-controlled drone swarms

Why xAI is throwing its hat in the ring after previously saying AI-powered weapons without a human in the loop should be banned. The development is part of the US's drive for drone swarming capabilities.

Ukranian soliders with drones were flying blind during the Starlink outage

SpaceX’s xAI is joining the contest to develop intelligent drone swarms for the United States. There seems to be no stopping the future of AI-enabled weapons of war with the power to kill autonomously, with some resigning to “if you can’t beat them, you may as well join them.”

SpaceX’s xAI to compete for DoD drone swarm contract

According to new reporting by Bloomberg yesterday, Elon Musk’s SpaceX and xAI subsidiary are competing “in a secretive new Pentagon contest to produce voice-controlled, autonomous drone swarming technology.”

Red Cat Black Widow drone
Photo: Red Cat

Bloomberg cited “people familiar with the matter” but did not provide any names or titles. The competition is for a six-month $100 million prize challenge that was launched in January.

The competition “aims to produce advanced swarming technology that can translate voice commands into digital instructions and run multiple drones,” according to the publication. Musk announced a few weeks ago that SpaceX and xAI will merge.

Bloomberg pointed out that Musk appears to have done a complete reversal of his previous position, saying that AI should not be used as “new tools for killing people.”

Before the war in Ukraine, the topic of “swarming drones” became something of a cliché and buzzword. Swarming is hard, very hard. AI technology paired with drones is increasingly seen as the frontier of future warfare.

What is known about the drone swarm programme

The drone swarm competition was launched jointly by the Defense Innovation Unit. That unit is focused on including the participation of Silicon Valley startups and the Defense Autonomous Warfare Group (DAWG).

GDU UAV-P300 drone in flight
Photo: GDU

DAWG was recently launched under the current Trump administration and is part of the US Special Operations Command. It is also partly continuing the work of the Replicator initiative launched during the Biden Administration, which seeks to produce many thousands of expendable autonomous drones.

Bloomberg says the effort will progress in five phases. It will start with developing software and progress to real-life testing. The aim is for the programme to be for offensive purposes, with a January announcement saying the human-machine interaction “will directly impact the lethality and effectiveness of these systems.”

Jiutian drone swarm launch render
Photo: CCTV

A new AI Acceleration Strategy released by the Pentagon in January spoke of “unleashing” AI agents for the battlefield, all the way from planning military campaigns to potentially lethal strikes.

Get the latest aerospace defence news here on AGN.

Why drone swarming is hard

To date, drones are not swarming in combat, and the famous Chinese drone shows also do not count as swarming in the military sense. Those are simple, pre-programmed, centrally controlled spectacles with rigid paths, a single point of failure, no decision-making, and operating in a permissive environment.

There have been drone swarm trials in the US, China, and Israel, but there is nothing mature.

In a military sense, drone swarming refers to groups of drones that share information, make decentralized decisions (without a single master controller), adapt in real time to losses, jamming, and changing targets, and coordinate effects.

The concept draws from swarming seen in nature, think flocks of birds, schools of fish, bee swarms, and ant colonies. Drone swarms should not have a single point of failure.

Swarming drones don’t just do the same thing; they coordinate with some scouting, others jamming radars, others striking, and others relaying information.

True miltiary swarming is extremely difficult. Overcoming jamming, operating in GPS-denied zones, electronically contested battlespaces, and developing robust algorithms for swarm intelligence is challenging. Bandwidth limits with many nodes sharing sensor data can saturate networks.

Photo: Ukraine MoD

Sign up for our newsletter and get our latest content in your inbox.

More from